THE ACCOUNTING REVIEW

Vol. 79, No. 4

2004
pp. 1153-1180

The Effect of Accounting Report

Structure and Team Structure on
Performance in Cross-Functional Teams
Casey Rowe
Arizona State University
ABSTRACT: Cross-functional teams operate in dynamic environments in which it is
difficult to properly align incentives. As a result the free-rider problem can emerge. This
paper presents two experiments in which dominant incentives to free ride were held
constant. The first experiment examined the question of whether aligning accounting
report structure and team structure in such a way as to create a "group frame" helps
to mitigate the free-rider problem. Accounting report structure was found to comple-
ment team structure. When properly aligned, accounting and team structures helped
to resolve the free-rider problem interactively by operating as a powerful group framing
device. The second experiment provides theory-consistent evidence that people out-
side the teams (subjects who acted as management control system designers) fail to
appreciate how powerfully the design of accounting and team structures influence per-
formance within teams. Finally, the study provides additional insight into the role that
accounting structure can play in promoting informal control within cross-functional
teams.

Keywords: accounting structure; cross-functional teams; free-rider problem; informal

control.
Data Availability: Data wilt be made available by the author upoti request.

I. INTRODUCTION

A
llhough cross-functional teams have become increasingly popular in practice, ac-
counting research has focused mainly on workgroups.' In contra.st to workgroups
that perform routine tasks within a single unit or responsibility center of the firm

' Fifly-six percent ot management accountants reported that they had worked on a cross-functional team in 1999,
a 17 percent increase from 1995 (Siegel and Sorensen 1999).

This study is from my dissertaiion. I gratefully acknowledge the many thoughtful comments from the members of
iny dissertation committee: Jacob Bimberg (Chair), Kathleen Carley, John (Hiirry) Evans 111, Donald Moser, and
Michael Shields. I appreciate the helpful comments from Terry Shevlin (editor) and two anonymous reviewers. [
thank David Cooper, Jeffrey Hales, Lynn Hannan. Vicky Hoffman, Steve Kacheimeier, Steve Kaplan, Lisa Koonce,
Joan Luft, Ed O'Donnell, Buck Pei. Mike Petersen, Joe Schultz. Nathan Stuart, Alisa Ti.sdale-Rowe, and seminar
participants at the 2(X)3 Management Accounting Research Conference, Arizona State University: University of
California, Riverside; Case Western Reserve University: Naval Po.stgraduate School: University of Pitt.sburgh; and
The University of Texas at Austin for their comments. This project was supported by a grant from the Institute
for Industrial Competitiveness.
Editor's note: This paper was accepted by Terry Shevlin, Senior Editor.
Submitted September 2002
Accepted September 2004

1153
1154 Rowe

(Young et al. 1993; Drake et al. 1999; Towry 2003), cross-functional teams perform non-
routine tasks that span the broader process or value chain {Wruck and Jensen 1994; Ittner
and Larcker 1995; Anderson et al. 2002). A central problem with cross-functional teams is
that they operate within a relatively severe control environment in which it is difficult to
properly align incentives.
Indeed, incentive related problems, such as free riding, are common within cross-
functional teams (Denison et al. 1996; Joyce et al. 1997; Scott and Tiessen 1999). Fortu-
nately, economic incentives are not the whole of the control package. The growing literature
concerned with trade-offs between formal economic incentives and informal control mech-
anisms demonstrates that social control mechanisms (culture, honesty, group identification)
can play an important role in mitigating agency problems (Kachelmeier and Shehata 1997;
Evans, Hannan, Krishnan, and Moser 2001; Towry 2003).
This paper presents two experiments in which dominant incentives to free ride were
held constant. Three questions are addressed. First: Does the informal control strategy of
aligning accounting and team structures in such a way as to create a "group frame" help
to mitigate the free-rider problem within cross-functional teams? Second: If creating an
appropriate alignment between accounting and team structures does help to resolve the
free-rider problem, then what informal control mechanism motivates team members to per-
form when they could have earned more by free riding? Third: If the configuration of
accounting and team structures does affect performance, then do people outside the teams,
who assume the role of control-system designers, appreciate how powerfully design alter-
natives affect motivation within teams?
Company-wide problems, such as process improvement initiatives, raise the question
of how to structure accounting reports and teams to best support the overall task (Simon
et al. 1954; Shields 1997). Articles in the popular press often presume that moving a cross-
section of managers (engineering, procurement, operations, etc.) together into "face-to-face
teams," from their geographically "distributed" locations throughout the firm, is sufficient
to induce high performance (Kiesler and Cummings 2002, 55). This paper presents a test
of the alternative hypothesis that achieving high performance depends on using the appro-
priate type of accounting report structure in combination with the appropriate team struc-
ture. Atkinson et al. (1997, 90) maintain that "there is great value in research that focuses
on the interaction between management accounting systems and work organization."
If the alignment between accounting and team structures does help to mitigate the free-
rider problem, then it is important to know whether people recognize the value of this
control mechanism. Designing effective management control systems depends on the ability
to predict how different design options influence performance (Sunder 2002). However,
behavioral theory posits that people outside the team have a biased understanding of how
strongly environmental factors, such as the way in which the management control system
is structured, influence performance within teams (Ross and Nisbett 1991).
Moreover, if creating an appropriate alignment between accounting and team structures
does influence performance, it is important to identify what informal control mechanism(s)
motivate team members to perform when they could have earned more by free riding.
Identifying how accounting systems influence social motives is vital to developing a more
complete picture of the role that accounting plays in solving control problems (Luft and
Shields 2003).
The first experiment examines the question of whether creating an appropriate align-
ment between accounting and team structures helps to resolve the free-rider problem within
cross-functional teams. In addition, if the configuration of accounting and team structures
does influence performance, then several competing and complementary informal control

The Accounting Review, October 2004

The Effect of Accounting Report Structure and Team Structure on Performance 1155

mechanisms are examined in an effort to explain why the way in which the control system
is structured affects the propensity to free ride. In a second experiment, subjects act as
management control system designers. The second experiment examines whether, given
strong incentives to make an accurate prediction, the designer-subjects can predict how
powerfully the control system design influences performance.
When properly aligned, accounting and team structures were found to operate inter-
actively as a powerful group-framing device. In particular, resolving the free-rider problem
depended on using process-level accounting reports in conjunction with face-to-face teams.
Consistent with the theory, designer-subjects failed to appreciate how powerfully the control
system design influenced performance. Designer-subjects also underestimated the severity
of the free-rider problem. Finally, consistent with the role of a "moral mediator" (Chua
and Degeling 1993), accounting helped to mitigate the free-rider problem by inducing both
trust and collectivism (an unselfish sense of duty to the team).
This paper contributes to the literature by demonstrating that accounting structure can
help to mitigate the free-rider problem indirectly by inducing informal control. However,
this only works if the cross-functional team structure as well as the report structure supports
group framing. The paper also demonstrates that people outside the team have a difficult
time appreciating the value of accounting and team structures in controlling free riding, as
demonstrated in experiment two. Finally, the study provides new insights into the role that
accounting structure can play in promoting informal control in a challenging cross-
functional team environment that can undermine standard formal control mechanisms.
The remainder of this paper is organized as follows: Section II focuses on the question
of whether and how the alignment between accounting and team structures can help to
resolve the free-rider problem. Section III presents experiment one and the results of this
experiment are reported in Section IV. Section V presents the design and results of exper-
iment two, which focuses on the question of whether designer-subject's appreciate how
strongly the structure of the management control system influences performance. Section
VI concludes with the discussion.

II. HYPOTHESES DEVELOPMENT

This section begins by developing an explicit model of incentives to free ride in cross-
functional teams. Next, research concerned with resolving the free-rider problem is re-
viewed. The following subsection presents the hypotheses.

Incentives to Free Ride in Cross-Eunctional Teams

I focus on cross-functional teams composed of unit-managers from several different
responsibility centers within the firm. Team members face the task of planning a process
improvement initiative (e.g., reengineering, TQM, just-in-time operations). I assume that
each manager possesses privately held specific knowledge about slack that is created by
the initiative. Revealing such private information is a direct cost to the individual because
it enables central management to identify and withdraw slack resources that could have
otherwise been "consumed.""-
Motivating unit-managers to reveal their specific knowledge can be problematic. The
task of adopting a process improvement creates high task interdependence between the

Stack is defined as ihe gap between specific knowledge of the potential cost savings from a proccs.s improvement.
and the actual cosi saving.s exiracled by the firm. For example, unit managers can consitme slack by exerting
less effort to meet standards or by purchasing additional perquisites (Antle and Eppen 1985).

The Accounting Review. October 2004

1156 Rowe

various parts of the firm that are itivolved (Teece 1996).' In tum, high task ititerdependence
makes causal responsibility for gains (improved performance) and culpability for los.ses
(free riding) indeterminate to some degree (Williamson 1975). Firms cope with this type
of ititerdependency problem by adopting team-level incentives that are contingent on joint—
rather than individual—performance (Bushman et al. 1995; Scott and Tiessen 1999).
In summary, the cost of revealing specific knowledge is borne by the individual, while
the benefit is shared. This cost-benefit pattern leads to a ubiquitous type of free-rider prob-
lem known as the public goods dilemma. This game is illustrated tiext. The same incentives
will be used in both experiments.
In the game, four unit-matiagers (n = 4) each possess "two units" of specific knowl-
edge (private information), costing the individual $5/unit to reveal. This is the opportunity
cost that arises from revealing slack and giving up resources. In an effort to motivate
contributions, the firm offers a team-level incentive of $8/unit, which is shared equally
(1/n) by the unit-managers. These parameters describe the public goods dilemma game
illustrated in Figure 1."*
Team performance is an increasing function of units contributed and a decreasing func-
tion of free riding (or social loafing).^ Figure 1 illustrates how a focal individual's payoff
depends on his or her own contribution (0-2 units) in combination with the sum of the
contributions made by the other three-team members (0-6 units). The individual incurs a
marginal cost of $3 (S8/4-$5) for each unit he or she contributes, while contributions by
other team members produce a marginal benefit to the individual of $2/unit ($8/4).
The dilemma is that team performance is higher and everyone may benefit if all four
"players" contribute fully; however, at the margin each individual is strictly better off free
riding (Dawes 1980). For example, when everyone cooperates fully, individuals earn $6
instead of $0, when no one contributes anything. Nonetheless, each individual can always
earn more by free riding. Conventional economic theory posits no one will contribute
anything when faeed with this type of situation (Ledyard 1995).

FIGURE 1
Payoff Matrix
Focal
Individuars Sum of Contributions from the Otber Three Team Members
Contribution 0 1 2 3 4 5 6
0 $0 $2 $4 $6 $8 $10 $12
1 ($3) ($1) $1 $3 $5 $7 $9
2 ($6) ($4) ($2) $0 $2 $4 $6

' Task inierdependence arises because advances in technology in .some areas of the economic system oHen cannot
be fully exploited or profitably ulilized unless a series of other connected tributary developments iKictir (Mitgrom
and Roberts 1992: Teece 1996). For example, an initiative to adopt Just-in-time operations can depend on
coordinated adjustmenis in engineering, purchasing, quality assurance, inventory cotitrol, manufacturing, and
accounts payable.
•" Subjects were endowed with $16 to ensure that they earned at least $10 for participating in the study. The
necessary condilions for the public goods dilemma are ihat the trulhtulness of an individuaFs revelation is not
directly observable (i.e., the information is subject to strategic misrepresentation) and that the costs of production
are born individually while the benefits are shared (i.e.. the team production problem). The sufficieni condilion
is that, at the margin. Ihe individual's cost of contributing exceeds his or her share of the benelils.
^ Free riding arises when one benefits without contributing one's share of the inputs to create the group's joint
output. Social loafing arises when all the team members withhold their potential contributions to an equal degree
(i,e,, everyone chooses to contribute either 0 or I unit out of 2 possible). To ease exposition. 1 use the term
"free riding" to refer to both free riding and social loating.

The Accounting Review, October 2004

The Effect of Accounting Report Structure and Team Structure on Performance 1157

Resolving the Free-Rider Problem

Research concerned with resolving tlie free-rider problem follows two distinct strate-
gies. First, n formal control strategy relies on the firm's ability to create incentive compatible
rewards to motivate goal congruent behavior (performing rather than free riding). In con-
trast, an informal control .strategy relies on the firm's ability to induce social control mecli-
anisms that motivate goal congruent behavior—in the absence of incentives that are prop-
erly aligned in the "formal"sense.
Forma] control strategies encounter several difficulties in the context of cross-functional
teams. First, as previously discussed, the task of adopting process improvements creates
heightened task interdependence. High task interdependence can confound standard forma]
contro] mechanisms such as transfer pricing schemes (Abdel-khalik and Lusk 1974, 8-9)
and truth inducing contracts (Waller and Bishop 1990). Seeond, mutual monitoring contracts
(Towry 2003) require that "agents' task over]ap (or duplicate) each other" (Holmstrom
1982, 338). This requirement is not met because cross-functional teams integrate "specific
knowledge possessed by different individuals performing traditionaUy quite separate cor-
porate functions" (Jensen and Meckling ]995, 7). Mutua] monitoring contracts also require
that team members work in close proximity so they can observe each other's output. How-
ever, members of cross-functiona] teams often perform their respective tasks in different
geographic ]ocations within the firm (Kies]er and Cummings 2002) and therefore direct
observation is not feasible. Fina]]y. although reputation may function as an effective check
on the free-rider problem when individuals enter into repeated transactions, reputation is
less effective in the nonroutine transactions that cross-functiona] teams typica]ly perform
(Mi]grom and Roberts 1992).
Informal control strategies rely on shared understandings (e.g., trust) to govern potential
opportunism (Birnberg and Snodgrass 1988; Chua and Degeling 1993: Sunder 2002). One
such informal mechanism is national culture. For example, Kacheimeier and Shehata (1997)
sliow that nationa] culture (in combination with the degree of anonymity in internal re-
porting) infiuences the level of free riding when teams face the public goods dilemma. The
extensive literature on public goods provision problems demonstrates that several other
informal .socia] factors can help to mitigate the free-rider problem. These factors include
allowing communication among the players, helping individuals understand whether their
contributions have a sizable impact on performance (efficacy), and framing the players as
belonging to a common group (see Ledyard [1995] for a review). I focus on this latter type
of informal contro] strategy.

Creating a Group Frame with Accounting Report Structure and Team Structure
This subsection develops the thesis that the way in which accounting report structure
is aligned with team structure can he]p to resolve the free-rider problem by creating a group
frame.
There is growing reeognition that the way in which accounting reports and/or teams
are structured is an important means of managing boundaries within the firm (Kilmann
1983; Jensen and [VIeckling 1995; Briers and Chua 2001).'' The concept of a boundary
refers to physical or conceptual relational limits that define entities (individuals or groups)
as separate from one another (Ashforth et al. 2000, 473-474). In particular, the absence

Economic architecture theory recognizes that team structure can imply how decision rights are .structured (Jenseti
and Meckling 1995, 14). However, economic architecture theory does not predict thai chaaging decision rights
alone is sufficient to motivate performance (creating appropriate performance measures and incentive compatible
rewards is also necessary).

The Accounting Review, October 2004

 1158 Rowe

(presence) of a boundary between people implies a group frame (individual frame). In turn,
the group frame (individual frame) affects informal rules of the game such that free riding
is less (more) appropriate (Lind and Tyler 1988; Luhmann 1995; Messick 1999).^
Accounting Report and Team Structures
Company-wide problems raise the question of how to structure accounting reports and
teams to best support the overall task (Simon et al. 1954; Shields 1997). Firms typically
use responsibility accounting that consists of classifying accounting information according
to the individual unit with primary responsibility and presenting a separate unit-level report
to each unit-manager (Kaplan and Norton 1996)."^ Unit-level accounting reports impute
responsibilities to managers as isolated individuals (Pick 1971). An alternative is to use
process-level accounting report structure that combines activities from different units fol-
lowing the value chain concept (Shields 1997). For example, process-level reports are used
in connection with activity-based costing, the balanced scorecard, and competitive bench-
marking (Kaplan and Norton 1996; Evans, Hwang, and Nagarajan 2001). In contrast to
unit-level reports, process-level reports share broader group-level information about the
value chain that extends beyond the unit-manager's individual area of responsibility.
Cookson (2000) demonstrated that when experimental subjects manually calculated
their team-level (individual-level) information performance increased (decreased), despite
dominant incentives to free ride. Cookson's (2000) study suggests that team-level infor-
mation increases performance in general, independently of team structure. In contrast, this
study examines whether the effect of report structure depends on team structure. In addition,
Cookson's (2000) study did not address whether accounting information can operate as a
group framing device^—when people receive processed accounting information (an ac-
counting report) with the appropriate calculations already performed.
Firms can choose to organize cross-functional teams in different ways. Before joining
a cross-functional team, members normally work in different geographic locations within
the firm (Kiesler and Cummings 2002). One popular approach is to use distributed team
structure in which cross-functional team members remain in their separate locations. For
example, distributed specific knowledge can be collected via email or by a liaison that
gathers the dispersed information. Another approach is to W%Q face-to-face team structure
that locates the cross-functional team members together in close proximity for a joint meet-
ing. Consistent with the theory of a group frame, research finds that the mere act of moving
people together, independent of communication, can improve performance within teams
(see Kiesler and Cummings [2002] for a review). However, the prior research has not yet
examined whether deriving the benefits from face-to-face teams depends on using process-
level accounting reports.

Examining the Interaction between Accounting Report Structure and Team Structure
When combined, process-level report structure and face-to-face team structure are ex-
pected to mitigate the free-rider problem by signaling the absence of boundaries betvi'een
team members, consistent with a group frame. Only the combination of process-level ac-
counting report structure and face-to-face team structure blurs the boundaries between the

' \ frame is defined as a cognitive representation that codifies expectations and assumptions about familiar events
(Minsky 1975). When a group frame is salienf. people "recognize that other people's outcomes or priorities
must sometimes be accepted and [that| their own desires must sometimes be delayed or toregone" (Lind and
Tyler 1988, 223). In contrast, when an individual frame is salienl. people focus more narrowly on their own
situation and they show less concern lor negative externalities they impose on others (Luhmann 1995. 266).
^ Unii-level reports "may include any measurable impact on other responsibility centers" (Bushman et al. 1995,
102).

The Accounting Review, October 2004

The Effect of Accounting Report Structure and Team Structure on Performance 1159

team members in a way that is consistent with a group frame. In constrast, other designs
are expected to exacerbate the free-rider problem by signahng the presence of a clear
boundary, consistent with an individual frame.''
Some combinations of accounting and team structures convey "mixed" cues about the
situation.'" For example, using either unit-level report structure and face-to-face team struc-
ture or process-level report structure and distributed team structure implies both the pres-
ence of a boundary and the absence of a boundary (or vice versa). Several lines of research
support the expectation that when people are presented with conflicting frames, the indi-
vidual frame dominates the group frame. First, when people are vulnerable to opportunism,
such as free riding, and they receive mixed signals about the trustworthiness of the others,
they tend to evaluate the low-trust signal (the individual frame) as being more credible than
the high-trust signal (the group frame) (Slovic 1993; Kramer 1999). Second, people tend
to anchor on the frame that is more salient and adjust insufficiently for the frame that is
less salient (Fiske and Taylor 1991). Gaertner et al. (2002, 585) find that an individual
frame is more salient than a group frame "due to ... availability, accessibility, and inescap-
ability of private feelings and goals." Thus, "mixed" management control system structures
are expected to induce low performance by conveying an individual frame.
Based on the above theory, only the combination of process-level report structure and
face-to-face team structure is expected to induce high performance by inducing a group
frame. In contrast, the alternative patterns of report structure and team structure (unit-level
report and face-to-face team, unit-level report and distributed team, or process-level report
and distributed team) are expected to induce low performance by implying an individual
frame. These expectations describe the perfectly ordinal interaction presented in Figure 2.

HI: Inducing high performance in cross-functional teams depends on using process-

level accounting report structure together with face-to-face team structure.
Examining How Accounting Report and Team Structures Affect Performance
If HI is supported, the question arises: Why does the alignment between accounting
report and team structures influence motivation to perform when team members could have
earned more by free riding? Identifying how accounting systems affect social motives is
important to developing a more complete picture of the role that accounting plays in solving
control problems (Luft and Shields 2003).
Prior research suggests that trust, group identification, and collectivism explain why
group framing affects performance. However, the prior research has been criticized along
several lines. First, the prior studies focus on a particular informal control mechanism
without attempting to rule out, or examine possible interactions with, alternative mecha-
nisms (Komorita and Parks 1995). Second, the prior studies rely on simple correlation
analysis that fails to test whether the informal control mechanism actually explains why
people perform instead of free riding (Dawes 1980; Messick 1999). Finally, the prior studies
generally rely on a single item measure of the behavioral mechanism and therefore construct
validity cannot be assessed (Kwok and Sharp 1998).
Trust is the willingness to increase one's vulnerability to others in the absence of
adequate safeguards (Zand 1997). In theory, group framing promotes "rapid." "situation-
based," or "depersonalized trust" when people first meet in temporary teams (Meyerson et

The presence of u botindary (or boundaries) is t;onsistent with ati individual frame bul not a group frame
(Kilniann 1983: Luhmann 1W5: Meyerson el al. 1995).
These "mixed" management conlrol system slruclures reHect the reality Ihm "[llarge organizations bundle
together many assets, and who has what decision rights may be ambiguous" (Milgrom and Roberts 1992. 290).

The Accounting Review, October 2004

U60 Rowe

FIGURE 2
Predicted Effect of the Alignment between Accounting Report Structure
and Team Structure on Performance

Performance;

Team Structure:
-•-Face-to-Face Team
-•—Distributed Team

Unit-Level Report Process-Level Report

Accounting Report Structure

This figure illustrates the predicted relation between accounting report structure, team structure,
and performance. The above illustrated pattern requires that (1) mean performance is high when
process-level accounting report structure is used in conjunction with face-to-face team structure and
(2) low under the other design alternatives. This pattern was used to select the contrast coefficients
for the planned contrast code model presented in Table 1.

ai. 1995; Kramer 1999). However, no matter how much others are trusted in the public
goods dilemma game, individuals can always earn a higher payoff by free riding. Thus,
trust alone is not sufficient to explain why people perform.
Kramer (1999) discusses that group identification or collectivism may explain how trust
induces performance. Distinguishing between group identification and collectivism is im-
portant because these informal control mechanisms operate differently with different im-
plications for the design of the management control system. Group identification is a type
of bias in which people contribute to their team's performance because they expect to derive
positive feelings of self-esteem from being associated with a successful group (Hinkle et
al. 1989). In contrast, all cultures instill a norm of collectivism that prescribes contributing,
rather than free riding, when one is faced with a group situation (Triandis 1995). Compared
to group identification, collectivism is a stronger, more binding, moral standard that is
"sustained by the feelings of embarrassment, anxiety, guilt and shame that a person suffers
at the prospect of violating [the norm] ..." (Elster 1989. 99).
The above theory leads to the following hypotheses that may not be mutually exclusive:

H2A: The alignment between accounting report structure and team structure induces
performance by inducing trust.
H2B: The alignment between accounting report structure and team structure induces
pertbrmance by inducing group identification.

The Accounting Review. October 2004

The Effect of Accounting Report Structure ami Team Structure on Performance 1161

H2C: The alignment between accounting report structure and team structure induces
performance by inducing collectivism.

III. EXPERIMENT ONE

The first experiment examined whether and how accounting report structure and team
structure can help to mitigate the free-rider problem in cross-functional teams. The single-
period experimental design was selected to simulate the nonroutine nature of the tasks that
cross-functional teams typically perform (Simon et al. 1954; Denison et al. 1996: Anderson
et al. 2002).

Subjects
Eighty-four M.B.A. students entered into the study while they were in the second month
of their program. On average the subjects had 4.4 years of business experience and they
had worked on 5.8 team projects in their jobs. Sixty-four percent of the subjects were male.
Forty-eight percent of the subjects listed their national culture as U.S. and 19 percent listed
their national culture as Asian.

Controlling Reputation
Efforts were made to simulate the temporary nature of cross-functional teams by con-
trolling reputation. First, the study was conducted using M.B.A.s who were at the beginning
of their program. Second, the experimenter ensured that subjects who had previously
worked together on a learning team (a prior M.B.A. program requirement) were randomly
assigned to different experimental teams. Finally, subjects were asked if they "knew" any
of the other members of their experimental team in the post-experimental questionnaire.
The mean result for "knew" was 3.5, which was below the neutral point on the scale (5),
toward "strongly disagree." This result provides evidence that reputation was experimen-
tally controlled.

Design
A 2 X 2 between-subjects design was used in which accounting report structure (unit-
level report, process-level report) and team structure {distributed team, face-to-face team)
were experimentally manipulated. The dependent variable was performance. The mediating
variables were trust, group identification, and collectivism.
Accounting Report Structure was experimentally manipulated by giving each of the
four players either a unit-level report showing only the individual's payoff under various
possible choices by the focal manager and the other three managers or a process-level
report with this same information plus process-level information summarizing the pay-
off for all four players under these combinations (Figure 3). The experimental reports
held constant the information subjects received. No new information was provided in
the process-level report. Subjects were informed of the simple "rule" used to compute the
reports (discussed in the procedure section). If they chose to do so, then subjects in
the unit-level report condition could easily compute all the information contained in the
process-level report.
Team structure was experimentally manipulated by organizing the four-team members
in either a distributed team or in a face-to-face team. In the distributed team experimental
condition, each team member was located in a different experimental room. In the face-to-
face team condition, the four members of each experimental team were seated together
around the same table. This manipulation was reinforced by referring to subjects as "unit-
managers" in the distributed team condition and as a "project team" in the face-to-face

The Accounting Review, October 2004

1162 Rowe

FIGURE 3
Accounting Report Structure
Unit-Level Report
Department-Level Report (Your Individual Payoff)
Units You Total Number of Units Invested By
Choose to Department B, C, and D Managers
Invest 0 1 2 3 4 5 6
0 $0 $2 $4 $6 $8 $10 $12
1 ($3) ($1) $1 $3 $5 $7 $9
2 ($6) ($4) ($2) $0 $2 $4 $6
Process-Level Report
Total Project Team Report
Number of
Units You Total Number of Units Invested By Other Three
Choose to Team Members in the Project
Invest 0 1 2 3 4 5 6
0 $0 $3 $6 $9 $12 S15 $18
1 $3 $6 $9 $12 $15 $18 $21
2 $6 $9 S12 $15 $18 S21 $24
Department-Level Report: {Your Individual Payoff)
Units You Total Numher of Units Invested By Other Three
Choose to Team Members in the Project
Invest 0 1 2 3 4 5 6
0 SO $2 $4 $6 $8 $10 $12
1 ($3) ($1) $1 $3 $5 $7 $9
2 (S6) (S4) ($2) $0 $2 $4 $6

team condition. No commutiication or overt signaling was permitted in any experimental

condition.
Performance was the dependent variable. Value-laden terms such as "cooperate," "re-
veal specific knowledge," or "free ride" were not used in the experiment. Rather the more
neutral term "invest units of resources" was used." Con.sistent with its definition as having
monetary value, specific knowledge was made operational by endowing each subject with
"two units of resources" worth $5 each. Performance increased with units invested and it
decreased with units withheld (free riding or social loafing).

IVust, Group Identification, and Collectivism

The Appendix presents the multi-item instruments used to measure trust, group iden-
tification, and collectivism. Trust was measured using two methods. First, subjects predicted
the number of utiits cotitributed by other team members. A $3.00 bonus was offered to

Experimental economic studies concerned wiih the publit; goods dilemma have avoided value laden lenns such
as "cooperate" or "free ride." instead using stark labels .such as " X " or "'O." The lenn "invest units of
resources" was used to avoid the value-laden alternatives while still providing a controlled degree of contextual
richness.

The Accounting Review. October 2004

The Effect of Accounting Report Structure and Team Structure on Performance 1163

motivate an accurate prediction (Ledyard 1995).'^ Second, trust was measured using a multi-
item instrument based on Kramer (1999). Group identification was measured using Hinkle
et al.'s (1989) instrument and collectivism was measured using an instrument based on
Triandis(l995).

Procedure
Subjects were informed that they would take part in a study on decision making that
required approximately 40 minutes to complete and that the average earnings in an earlier
(pilot) study had been approximately $20. Upon entering the experimental facility, each
subject signed an agreement not to talk except to ask the experimenter a question. Subjects
also agreed not to use hypothetical examples in any questions they asked. A penalty of
immediate dismissal with only the $5 show up fee was imposed for violating the rules.
None of the rules were violated in the experiment, and therefore, the penalty was never
invoked."
After randomly assigning subjects to treatments and experimental cross-functional
teams, the initial materials were handed out and the experimenter read the instructions
aioud. Subjects were asked to assume the role of a unit-manager who works on a joint
project with three other managers from three other units of the XYZ Co. Each individual
was given an initial endowtiient of $6 and two units of resources worth $5 each {total
endowment $16). The experimental game was identical to the one previously illustrated in
Figure I. The subjects were fully informed about the nature of the game. The instructions
explained that the reports were constructed according to the following rule: "... each unit
of resource you choose to invest in the joint project costs you $5 and it earns $8, which is
divided equally among you and the three other managers who work on the joint project."
The experimenter worked through four examples and a quiz was administered to ensure
that all the subjects understood the game. The specific examples, as well as their sequence,
were held constant across both experiments. The quiz was administered to test subjects'
understanding of the incentives, including the "rule" used to calculate their reports. In the
event of an incorrect answer the pertinent explanation was reiterated. The experimenter did
not proceed until each subject answered all the questions correctly and he was convinced
that everyone understood the game and how the reports were computed.
In an effort to simulate the private nature of specific knowledge the experimental pro-
cedure was designed to create a high level of anonymity. This procedure also helped to
control for reputation and experimental demand. Subjects were informed that:
the experimenter and the other unit-managers will not know what choices you make
or how much money you actually earn. You have been assigned a random code on a
slip of paper that is your receipt to claim your pay. Your pay will be placed in a sealed
envelope and given to an independent cashier who will give you your envelope in
exchange for your receipt.
After being assured that no deception was involved and that the other players received
exactly the same information, each subject made his or her investment choice (0. I. or 2

Trust was measured after subjects choose how many units of resources to invest in the joint project but before
they learned uf their payoff.
The penalty was designed to prevent subjects from using their questions strategically to influence other team
members. This was done because two of the 24 subjects in the pilot sludy had used hypothetical examples to
signal the investment choice they wanted other team members to make. In debriefing, pilot-subjects said these
examples had influenced their investment choice.

The Accounting Review, October 2004

1164 Rowe

units). The experimenter ensured that no overt communication occurred. Individuals then
sealed their response form in an envelope and passed it to the experimenter. After making
their investment choice, subjects predicted the actual number of units (0-6) invested by the
other team members. Subjects were informed that they would receive a $3 bonus for a
correct prediction. Finally, the post-experimental questionnaire was administered. The day
after their experimental session, subjects claimed their earnings by presenting their coded
receipt to the independent cashier.'"* Subjects were debriefed via email shortly after the
study was completed.

IV. RESULTS OF EXPERIMENT ONE

The first experiment examined the effect of the alignment between accounting report
structure (unit-level report, process-level report) and team structure (distributed team, face-
to-face team) on performance when cross-functional team members faced dominant incen-
tives to free ride. Several informal control mechanisms were measured in an effort to explain
hypothesized performance effects.
All of the subjects completed the quiz correctly, providing evidence that they under-
stood the incentives in the game that they faced. Ninety-four percent of subjects correctly
answered a post-experimental question about the unit-level report or the process-level report
they had received.'*^ Ninety-eight percent of subjects correctly identified whether members
of their project team were located in other experimental rooms (distributed team condition)
or in seats around their table (face-to-face team condition). The results of the manipulation
checks did not vary across experimental treatments. Based on these results the experimental
manipulations were considered successful.
Subjects were successfully randomized to the experimental treatments. National culture
was randomized effectively and it did not affect the results."" The statistical assumption of
independence might have been violated if nonverbal communication had occurred when
players were in the face-to-face team condition. The hypothesis of non-independence was
tested following Kenny and Judd (1996).'^ The null hypothesis of independence was not
rejected (t = .62, p = .27, one-tailed). Therefore, individual subjects, rather than experi-
mental teams, were used as the unit-of-analysis for statistical purposes.
The first hypothesis stated that achieving high performance depends on using process-
level report structure together with face-to-face team structure. All other patterns of report
structure and team structure were expected to result in low performance. This interaction
(Figure 2) was tested using a planned contrast model."* Figure 4 presents the frequency of
individual and team-level performance and Table 1 presents the test of HI.
Consistent with HI. performance was higher when the process-level report was used
in conjunction with the face-to-face team (mean = 1.13 units, 56.5 percent of the total)
compared to the average of the other report structure and team structure experimental
conditions (mean ^ 0.63 units, 31.5 percent of the total) (t = 7.30, p < .01, one-tailed).

''' The independent cashier did not know [he purpose of the study.
'•^ One subject did not answer either of the manipulation check questions.
"• Unhke Kachelmeier and Shchata (1997) the experiment was not designed to detect effects of national culture
(a larger sample size is needed to separate cultural variation from individual varialion).
'^ This statistical test compared the covariance of the dependent measure (pertbmiance) within experimental teams
to the covariance of the dependent measure across experimental teams {Kenny and Judd 1996).
"* Buckless and Ravenscroft (1990, 935) demonstrate that although the default weights assigned in the conventional
ANOVA provide a powerful test of disordinai interactions. "ANOVA is less powerful as a statistical tool for
testing ordinal interactions." They go on to recommend that when researchers expect an ordinal interaction a
priori, "they should rely on contrast ccxling instead of conventional ANOVA" (Buckless and Ravenscroft 1990,
937).

The Accounting Review. October 2004

The Effect of Accounting Report Structure and Team Structure on Performance 1165

FIGURE 4
The Effect of Accounting and Team Structures on Performance—Hypothesis 1

Performance
Unit-Level Report Process-Level Report Unit-Level Report Process-Level Report
~ and Distributed Team and Distributed Team and Face-to-Face Tearn and Face-to-Face Team

(n = 20} (n = 20) {n = 20) (n = 24)

I 1
r 0
-
1 11 11 U-Lj-

(Sorted by rate of contributor!)

Contributions by a Four-Person Experimental Team (rartge 0 - 8 units}

Contributions by Individuals (range 0-2 units)

Panel B in Table 1 presents the results of the contrast eode model testing whether using
the proeess-level report together with the face-to-face team structure causes high perform-
ance. The result of this test provides support for HI (F = 6.41. p < .01, one-tailed). Also
consistent with HI, the other experimental cells (unit-level report structure and face-to-face
team structure, unit-level report structure and distributed team structure, or process-level
report structure and distributed team structure) were not statistically different (F = .39, p
= .68, two-tailed). In summary, achieving high performance depended on using process-
level report structure and face-to-face team structure simultaneously, while only low per-
formance was achieved under the alternative report and team structures.

How Do Accounting Report and Team Structures Help to Mitigate the

Free-Rider Problem?
Recall that trust (H2A), group identification (H2B), and collectivism (H2C) may explain
how the alignment between accounting report structure and team structure influences per-
fomiance. The first measure of trust was the number of units other team members were
predicted to contribute to the joint project (0-6 units). Panel A in Table 2 shows the mean
number of units other team members were trusted to make and Panel B presents descriptive
statistics on the differences between predicted and actual units contributed.
Based on the data reported in Panel A of Table 2, on average, subjects trusted other
experimental team members to contribute significantly more when the process-level report
was used in conjunction with the face-to-face team structure (3.75 units or 62.5 percent of

The Accounting Review. October 2004

1166 Rowe

TABLE I
The Effect of Accounting and Team Structures on Performance—Hypothesis 1

Panel A: Descriptive Statistics"

Accounting Report
Structure
Unit-Level Process-Level
Report Report
Team Structure 1 2
0.75 1.13
Face-lo-Face Team (0.64) (0.80)
1201 [24]
{-11 [3}
3 4
0.55 0.6
Distributed Team (0.89) (0.68)
[20] [20]
{-1} {-11

Panel B: Test of Hypothesis 1 that Cell 2 > Cells 1, 3, and 4''

SS df MS E-stat Significance
Contrast Model 3.70 I 3.70 6.41 <.OI, one-tailed
Model—Unexplained SS .87 2 ,44 .732 .395
Total 4.57 3 1.52 4.33 .01

Statistical Equivalence of Cells 1, 3, and 4

SS df MS E-stat Significance
Semiomnibus F-Test .43 2 .21 .392 .677, two-tailed
Error 31.5 57 .55

^ Panel A reports perfonnance out of a maximum of 2 units. Cells conlain means, (standard deviations), [number
of observations], and {contrast code|.
''The contrast codes used are the ones suggested for use by Buckless and Ravenscroft {1990, 935) when the
prediced "etfecl is due entirely lo the difference of one experimental cell from the other ihree cells, which are
approximately equal" (see Panel A jcontrasl code}).

the total) compared to the average of the other experimental cells (2.88 units or 48 percent
of the total) (t = 2.87, p < .01, one-tailed).'" Moreover, consistent with H2A and the
expectation that trust motivates performance, if individuals chose to calculate it, their av-
erage expected payoff from contributing fully (i.e., investing two units) was larger in the
former experimental cell than in the latter experimental cells. Recall that the individual
incurs a marginal cost of $3 for each unit invested in the joint project while units invested
by other team member's produce a marginal benefit to the individual of $2/unit. Therefore,
the average expected payoff from investing two units in the process-level report and face-
to-face team condition was $ +L50 ($2 X 3.75 - $3 x 2) as opposed to only $ -0.24
($2 X 2.88 - $3 X 2), on average, in the other experimental conditions. These results

Predictions in the other experimental treatments (unit-level report and face-to-face team, unit-level report and
distributed team, and process-level report and distributed team) were not statistically different (F = .56, p
= .57, two-tailed).

The Accounting Review, October 2004

The Effect of Accounting Report Structure and Team Structure on Performance 1167

fl b.

2 « O O O —' —

liL

a-o o ^
_ s
o
a
>^
?> * _
?54E O
o o o
ON O
oj u n
X 3 ~

s e
a C CO
U
V

V
n
«-g
__ —t g
o Q r~ g
v 01 n d " d d 2
> W ,4)
5 u o
-Q
.•s -a
c c
o 5.2
ion

u
u
01
o 3
•a "2 >
U 3 s y
Li
•
ion ind

0S o q
U n. C
rate

1/^ —; V O
>
a.
3 -a
a -
a c
•S
C '.j o I.£
o tU
C c C
0£)
£ If.
• _ _

3 c
C i
0 u
XI u

B y

CQ

Accounting Review. October 2004

1168 Rowe

illustrate that trust can produce a substantial increase in the payoffs individuals expect from
engaging in joint production.
Panel B in Table 2 shows that predicted contributions by others were relatively accurate
across each of the experimental treatments. None of the prediction errors were statistically
significant. Only the proportion of subjects who earned a bonus for prediction accuracy
differed (unexpectedly) across treatments. Overall, these results indicate that subjects made
accurate predictions.
Composite measures of trust, group identification, and collectivism were computed
using factor analysis. Each of these constructs met acceptable reliability levels, with
Cronbach alphas between 0.78 and 0.85 (Kwok and Sharp 1998, 141). Two items from the
trust instrument had eigenvalues greater than l.O and they were weighted by their factor
scores to compute a single measure of trust. These items were the number of units other
players were predicted to invest (discussed above) and "I expect that other team members
considered my interests when they made their investment choice." Two items from the
group identification instrument had eigenvalues greater than 1.0 and they were weighted by
their factor scores to compute a single measure of group identification. These items were
"I think this team will work well together" and "I do not fit in well with the other members
of this team" (reverse scored). Finally, one item from the collectivism instrument had an
eigenvalue greater than 1.0. Therefore, the item: "I felt that I had a duty to invest my units
of resources in the team project" was used to measure the collectivism construct.
Figure 5, Panel A presents a test of the hypotheses that trust (H2A), group identification
(H2B). and/or collectivism (H2C) explain how the alignment between accounting and team
structures affects performance. The.se hypotheses were formally tested using Holmbeck's
(1997) four criteria to establish a mediating relationship. Figure 5, Panel A shows that only
trust fully satisfied each of Holmbeck's (1997) four criteria. In contrast, group identification
and collectivism achieved only marginal significance. Although trust met each of the nec-
essary and sufficient conditions to establish it as a mediator, trust did not completely explain
performance. The alignment between accounting report structure and team structure still
had a marginally significant effect on pertbrmance after incorporating trust as a mediator
(F = 2.57. p < .057, one-tailed).-" Neither group identification nor collectivism helped to
account for this residual effect. The unexplained variation could be attributable to mea-
surement error or to some other informal control mechanism that was not measured in this
study (Hays 1994, 836-837).
Recall that trust alone cannot explain why people perform when they could have earned
more by free riding. As Kramer (1999) discusses, group identification or collectivism may
explain how trust uffects performance. Accordingly. Figure 5, Panel B presents an analysis
of whether group identification or collectivism explains how trust infiuences performance.
Once again, Holmbeck's (1997) four criteria were used to evaluate the.se competing hy-
potheses. Panel B shows that only collectivism (H2C) mediated or explained the relationship
between trust and performance. However, some of the tests of group identification were
marginally significant and therefore group identification cannot be ruled out with much
confidence as an explanation for the group-framing effects. Overall, these results indicate
that both trust (H2A) and collectivism (H2C) are important informal control mechanisms
that are powerfully affected by the structural design of the management control system.

-" Holmbeck (1997) discusses that il would be unu.sual for a mediator to reduce the effect of the iiidependent
variables on tbe dependent variable from significance to zero.

The Accounting Review. October 2004

The Effect of Accounting Report Structure and Team Structure on Performance 1169

Analysis of Strategies
Additional analysis was conducted to obtain deeper insight into the question of how
the control system structure affects performance. In particular, five types of strategies were
inferred based upon the relationship between the individual's own contribution (0-2 units)
and the level of contribution he or she expected from the other team members (0-6 units).
These strategies were: conventional economic, free riding on others, contributing despite
expected free riding by others, trusting, and reciprocity. Table 3 presents the definition of
each strategy and the data.
Table 3 shows that regardless of the management control system design that was used,
most people did not perform unless they trusted others to forgo their opportunity to free
ride (expected "free riding ^v others" was less than or equal to 5 percent across the
experimental conditions). However, only a minority of subjects (15 to 25 percent) followed
a strategy consistent with reciprocity. Consistent with the previous analysis, trusting strat-
egies were more common when the control system structure was aligned with a group
frame (75 percent versus 45 percent, on average in the other experimental cells). In addition,
group framing resulted in a lower incidence of self-interested economic strategies (4 percent
versus 13 percent, on average in the other experimental cells) and a lower propensity to

FIGURE 5
Alternative Explanations for Performance

Panel A: Examining How Accounting and Team Structures Affect Performance—Hypotheses 2A, 2B, and 2C

Independent ia) Direct Effect Dependent

Variables (IVsl: (d) Decline in Direct Effect Due to Mediator Variable (DVl:
Accounting & Performance
Team Structures

i Face-lo-Face Teai"
* H2A Trust -^
0( Other Skucluresi H28 Group Identitication
H2C Collectivism

This figure presents a test afH2A,H2B. and H2C that trusi, group identification,and/or collectivism explains why the design of accounting and leam
structures affects pefformance. These tiypottieses were lasted using Ihe following four criteria specjiied by
Holmbeck (1997, 599) to testa mediating relation'

Hypotfieses Tests to Establish a Mediating Relation Between Accounting and Team Structures and Performance
Potential Mediators:
(a) the IVs effect the DV F = 7.30, p < .01 v' ' H2A Tfust H2B identification H2C Coiiectivism
(b) the IVs are associated with the mediator F ^6.0B,p < .01 F = 1.e5, p = ,09 - F = 1.32,p^.13 ~
(c)ttie mediator associated with the DV = 5 99. p < 001 f =2.56, p <.O1 ^V ( =6.79,p <.mj
(d) the mediator causes B\e direct effect (a) to decline F = 4,73,p < .05 V F = 1.51,p=.11 - F = 1.99, p = .09 ~

The analysis shows that only trust achieved significance (deruled by "^j for each of Ihe necessary and sufficient criteria to establish a
mediating relationship, while group identification and colectivism achieved only marginal significance on criteria (b) and (d) (denoted by ~)

"Specifically, the F-staiistic and one-tailed p-value associated wilh path (a) is the 1 X 2 ANOVA with performance as ihe
dependent variable and accounting & team structures as the IV. The F-statistic and one-lailed p-value associated with path
(b) is the I X 2 ANOVA with ihe mediator (trust, group identification, or collectivism) as the DV and accounting & team
structures as the IV. The F-siaiistic and one-tailed p-value assix^ialed with path (c) is from a linear regression with
performance as the DV and the mediator (Irusi, group ideniificaiion, or collectivism) as the IV. Finally, the F-statistic and
one-lailed p-value associated with path (d) is the change in the significance of paih (a) after entering ihe mediator (trust
group identification, or collectivism) as a control variable in the 1 X 2 ANCOVA with perfonnance as the dependent
variable and accounting & team structures as ihe IV,

{continued on next page)

The Accounting Review, October 2004

1170 Rowe

FIGURE 5 (continued)

Panel B: Examining How Accounting and Team Structures Affect Performance—Hypotheses 2B versus 2C

Dependent Variable (DV):

Accounting &
Performance
Team Structures
f Process- Level Repod
& Face-to-Face Team
or Othei MCS Structures)

^ H2B Group Identification vs.

H2C Collectivism

TTiis panel presents a test of whether group identification (H2B) or collectivism (H2C) explains how accounting and team
structures affect trust and pertomiance. These alternatives were tested using Holmbecks (1997, 599) four criteria:

Hypotheses Tests to Establish a Mediating Relation Between Trust and Performance

Potential Mediators:
(a) the IV effects the DV ( ^ 5,99, p < .001 \' H2B Identification H2C Coilectivism
(b) the IV is associated witfi the mediator I - 474. p < .001 V (=6.13,p<.001
(c) the mediator associated with the DV f = 2.56, p = .06 - f = 6.79, p < .001
I = ,203. p - ,42 n,s. f ^ 4.24, p < 001
(d) the mediator causes the direct effect (a) to decline

The analysis shows that only colleclivism achieved significance (denoted by V ) for each of the necessary and sufficient criteria to
establish a mediating relationship between trust and performance. As in the prior Panel, ~ denotes marginal significance. T ^ e n
together, the results presented in panels 5A and 5B indicates that accounfing and team strucftjres influence performance by affecting
trust (H2A} in combination with collectivism (H2C).

statistical analysis follows the same steps as described in panel ."ia with the following exceplion: linear regression was
used (rather than ANOVA) lo test ilenis (a), (b). and (d). All uf the reported p-values arc one-tailed.

free ride <)n others (21 percent versus 39 percent, on average in the other expetHmental
cells). In summary, using process-level report structure together with face-to-face team
structure seems to motivate performance by inducing trust and by making free riding on
other team members less appropriate.

V. EXPERIMENT TWO
Do people appreciate how powerfully the design of the management control system
influences performance within teams?
Creating an effective management control system depends on the ability to accurately
predict how different types of designs influence performance {Sunder 2002). Theories of
control rely on designer's skill to recogniz-e a ''misHt'" and redesign the management control
system (hereafter MCS) to better "fit" the environment (Donaldson 2001). However, be-
havioral theory supports the expectation that people fail to recognize how strongly alter-
native MCS designs influence performance. The second experiment .shifts the focus from
subjects embedded within cross-functional teams (experiment one) to subjects who act as
control system designers. Experiment two examines whether, given strong incentives for
accuracy, designer-subjects can predict how powerfully different patterns of MCS structures
influence performance within teams. Experiment one provided the data about how the team
members actually performed and this data enables the measurement of prediction accuracy
by designer-subjects in experiment two.

The Accounting Review. October 2004

The Effect of Accounting Report Structure and Team Structure on Performance 1171

at at
at 5 «
K: J3 «
-J § ^ E
— o
if s^
•^ £
C 3

X
•c
3 Xl
E

rib
M >. a,
k. oj a>
o u o
at h t., Q c= E <n
i 0- U= o
a o 0) o
E £ t~
u-
cd
<U >. 8 O O
cr

roci
lU LU
XI XJ Ul
o <u O 51
E

rat
U
E D. O
Ol
c c:
a.
a

nti
X ^ a,
cc °c .^
> il> en
<u
a>
> -a
B i O
o V
U-
«) =)
1 1
Si

D
othe
o o
fr

c a. P 2 ^ a.
o —u „
U fc ^
CD CT)
'c ox: •£
i3 2^
St

at

*- 4t .2i E
X •
c Sum of Contributions from
Other Team Members
UD
«S E tii - o
CQ ^ tu ^ to C\J
< a.
at b a«
2 lu
-I •B
3
Xt
•r- o
^1 •a
V
CO 0
a po

^E
ed

n
w o C\J "* ^ CVI
E S
o
a> . 2
5 3
CM T
— 3
•fc X I cd - ^
" a i , 3 CO t o
X)
•n li 0 0

c
O C O O c og Cd C
u. O
Contribution:

O .IC C
Individual's

u- O S o O C u O
.a
Focal

•- -o o -a
c c 60 bu CD eg

-S '£
3 3 fD P C
C ^ O^ S8 lA Xi

SE
c

yi i/j 5-5 — <U (U 01

^^x|
ill
soo
•y,

ex "Z
W c O
«:: u
u U
Oo ^ O
§ ^^ E
^ OJ

tu c >. c OXJ D H
-5 2£.
60 00 > _o c e
c c
•5 y
ii .a 00
^o 2J 5-
c/:
OJ
c jo Jo
x'S. s
s
t5 5 s X) £ 'Oij >
OJ c
o
a* Oi
a
aj OJ ^ ra U
g £ iJ Ji E

The Accounting Review. October 2004

 172 Rowe

Much of the management control literature assumes that control system designers can
avoid a "misfit" by identifying the design that best "fits" the environment. For example,
Donaldson (2001. 257-258) concludes that the "contingency literature seems to take it as
largely unproblematic that when an organization is in misfit, its management can see what
would be a tit." In contrast, the growing literature on "resistance" to management account-
ing change highlights difficulties designers have in selecting appropriate MCS designs (e.g.,
Anderson et al. 2002).
The behavioral theory of the fundamental attribution error predicts that people who are
located outside teams fail to recognize how strongly the way in which the MCS is designed
actually influences performance within teams. This theory explains that the actors who
work within teams are more acutely aware of contextual or environmental factors (MCS
structure, incentives to free ride) and that these factors influence the actor's performance
more powerfully than observers (those outside the team) would have predicted (Fiske and
Taylor 1991; Ross and Nisbett 1991). This bias may arise from a "simulation failure," in
which the observer's actual current perspective (from the outside looking in) clouds their
ability to imagine the hypothetical situation the actor's face (Blount and Larrick 2000).
A long line of research documents the bias predicted by the fundamental attribution
error (see Fiske and Taylor [1991] for a review). More recently, this line of research has
begun to examine organizational settings. Blount and Larrick (2000) demonstrated that
when negotiators chose the bargaining context that their bargaining partner would face, they
failed to choose the context that had been shown to induce a favorable outcome. As a result,
those who chose the context earned lower payoffs than if they had used the context to their
full advantage. In a different setting, Weber et al. (2001) found that, despite incentives for
accuracy, people who observed others playing a coordination game underrated how strongly
changes in an environmental factor (team size) actually affected the player's performance.
However, the prior research has not examined whether observers recognize how MCS struc-
tures and incentives to free ride affect performance within teams.
Based on the above theory, designer-subjects are expected to underestimate the mag-
nitude of the free-rider problem. In addition, the theory supports the expectation that
designer-subjects fail to recognize how strongly the use of alternative MCS structures affect
performance within teams. This pattern of expectations is presented in Figure 6 and below
in H4.

H3: Designer-subjects underestimate the severity of the free-rider problem.

H4: Designer-subjects underestimate the performance impact of using process-level re-
port structure together with face-to-face team structure versus unit-level report
structure together with distributed team structure.
Method: Experiment Two
Subjects, Design, and Task
Forty evening M.B.A.s were recruited to act as designer-subjects in the second study.
The evening M.B.A.s were chosen because they had a relatively high level of experience
managing others. Sixty-five percent of the subjects were male. The subjects were informed
they could earn between $5 and $20 for participating in an experiment that would take
approximately 15 minutes. The subjects were told that their earnings would depend on their
ability to accurately predict the actual performance of another group of daytime M.B.A.s
(those who assumed the role of cross-functional team members in experiment one). Ex-
periment two used a 1 X 2 between-subjects design. After being fully informed about the

The Accounling Review. October 2004

The Effect of Accounting Report Structure and Team Structure on Performance 1173

FIGURE 6
Experiment 2: Predicted versus Actual Effect of Accounting and
Team Structures on Performance—Hypothesis 4

Performance

Process-Level Report
and Face-to-Face Team

Unit-Level Report
and Distributed Team

Predicted Actual
(Designer-Subjects (Cross-Functional Team-
in Study Two} Subjects in Study One)

Planned Contrast

Predicted vs. Actual

This figure illu.slratcs the expecled relalion betweeti the alternate MCS designs and petformance. The above
illuslraied paltem requires that designer-subject.s undere.stimate the atual impacl of using alternate actroLnting and
team structures on pcrfomiiince. This pattern was used to select the contrast coefficients for the planned contrast
code model used to test H4.

incentives used in the first experiment (Figure I), designer-subjects estimated the cross-
functional team member's performance under either process-level report structure and face-
to-face team structure or unit-level report structure and distributed team structure. Recall
the unit-manager-subjects had to choose whether to contribute zero, one, or two units.
However, the average could reflect partial units. Designer-subjects predicted this average
on a 0-2 scale that was divided into 0.1 increments. In addition to their $5 show-up fee,
subjects received a $5 bonus if their prediction was within ±0.1 increment of the average
or a $15 bonus if their prediction was correct.

The Accounting Review. October 2004

1174 Rowe

Procedure
The designer-subjects were provided with a complete description of the game used in
experiment one. The experimenter worked through four examples and a quiz was admin-
istered to ensure that all the subjects understood the game. The specific examples, as well
as their sequence, were held constant across both experiments. The wording used to describe
the game and the experimental manipulations was also held constant. After the quiz was
corrected and the questions were answered, designer-subjects read a complete description
of either the process-level report and face-to-face team experimental treatment or the unit-
level report and distributed team experimental treatment used in study one. They then
predicted the average level of performance.

Results Study Two

Six subjects earned a $5 bonus for predicting the average level of performance within
±0.1 of a unit (evenly split between the process-level report and face-to-face team treatment
versus the unit-level report and distributed team treatment). One subject earned the $15
bonus for making an accurate prediction (unit-level report and distributed team treatment).
Hypothesis 3 predicted that designer-subjects would underestimate the magnitude of the
free-rider problem. Figure 7 presents the predicted-level of free riding by designer-subjects
in study two (mean of 0.88 units, 44 percent of the total possible) versus the actual level
of free riding by unit-manager-subjects in study one (mean of 1.16 units, 58 percent of the
total possible). Consistent with the theory of the fundamental attribution error, designer-
subjects underestimated the severity of the free-rider problem (F = 3.13, p < .05. one-
tailed). These results provide support for H3.
Hypothesis 4 posited that designer-subjects would underestimate how powerfully alter-
native MCS structures would influence performance within teams. Although performance
differed significantly in experiment one when either process-level report and face-to-face
team structures or unit-level report and distributed team structures were used (difference of
0.58 units, 29 percent of the total possible) (F = 5.12, p < .01, one-tailed), designer-
subjects failed to fully predict this effect in study two (difference of 0.27 units, 13.5 percent
of the total possible) (F = 1.93, p = .09, one-tailed). Figure 7, Panel B presents the results
of the planned contrast test of H4. The results provide support for H4. Designer-subjects
underestimated the performance impact of using process-level report structure and face-to-
face team structure versus unit-level report structure and distributed team structure (F
= 10.50. p < .01). Overall the results are consistent with the theory. Designer-subjects
failed to recognize how strongly incentives to free ride and different types of MCS struc-
tures affect performance within teams.

VI. DISCUSSION
This paper presents two experiments in which incentives to free ride are held constant.
The first experiment examines the question of whether aligning accounting and team struc-
tures in such a way as to create a "group frame" helps to mitigate the free-rider problem
in cross-functional teams. In addition, the first experiment examines the question of how
(if at all) accounting and team structures affect motivation to free ride. The second exper-
iment examines the question of whether subjects outside the teams, who act as management
control system designers, appreciate how powerfully the design of accounting and team
structures infiuences performance within teams.
Experiment one demonstrates that, when properly aligned, accounting and team struc-
tures operate interactively as a powerful group framing device that helps to resolve the
free-rider problem. Consistent with the theory, the second experiment demonstrates that

The Accounting Review. October 2004

The Effect of Accounting Report Structure and Team Structure on Performance 1175

FIGURE 7
Designer-Subject's Ability to Predict tbe Impact of MCS Design on Performance

Panel A: Predicted versus Actual Effect of MCS Design on Performance—Hypotheses

3 and 4

Mean Free Riding: 1.16

Performance

1.4- Process-Level Repoftand

1.26 Face-to-FaceTeam (MCS 1)
1.2- 1,12 1.13
Mean (Average of MCS 1
10- .99
and MCS 2)
.84
0.8- Unit-Level Report and
Distributed Team (MCS 2)
0.6- .55

0.4-

0.2-

0 - 1
Predicted Actual
(DesJgner-Sjbjects (Cross-Functional Team-
in ExperimentTwo] Subjects In ExperimeitOne}

This figure reports free riding and performance out of a maximum of 2 units.

Panel B: Test of Hypothesis 4: Predicted (MCS 1 - MCS 2) < Actual (MCS 1 - MCS

SS df MS F-Stat Signif.

Contrast Model 5.70 1 5.70 10.50 <.0l

Model-Unexplained SS 0.18 2 0.09 0.33 >.1O
Total 5.88 3

'' Mean free riding was calculated by subtracting mean performance from tbe maximum level of 2 uniis.
^ The contrast codes used are the ones suggested by Hays (1994, 427) "to compare the mean of the first two
groups with the mean of the last two groups."

The Accounting Review. October 2004

1176 Rowe

designer-subjects fail to appreciate how powerfully the design of the accounting and team
structures influence performance. Designer-subjects also underestimated the severity of the
free-rider problem. Finally, accounting and team structures were found to help mitigate
the free-rider problem indirectly by inducing both trust and collectivism (an unselflsh sense
of duty to the team).
In experiment one, subjects assigned to the unit-level accounting report condition had
to perform additional calculations to produce information that was provided in the process-
level report condition. Therefore, cognitive laziness could explain why report structure
affects performance. However, cognitive laziness cannot account for the interaction that was
observed between report structure and team structure. Although trust and collectivism
largely explained how accounting and team structures influenced performance, a marginally
significant effect remains unexplained. Follow-on research can contribute by examining
other informal control mechanisms and by improving measurement reliability.
Several factors may limit the ability to generalize the results of this study to practice.
First, the subjects were endowed with monetary resources to proxy for their specific knowl-
edge. People often value resources given to them differently than resources they had worked
to obtain (Boylan and Sprinkle 2001). Second, the way in which specific knowledge was
made operational did not require subjects to lie in order to withhold their private infor-
mation. Lying may change the nature of the game. For example, Evans, Hannan, Krishnan,
and Moser (2001) find people experience disutility from making dishonest revelations.
Third, team members knew how many units of resources each other possessed. Thus, the
private nature of specific knowledge was not fully simulated in this study. Finally, certain
parameter values were selected. The experiments did not examine how sensitive the findings
are to changes in the parameters used.
This paper contributes to the literature by demonstrating that accounting structure can
mitigate the free-rider problem by fostering informal control. However, accounting by itself
is of little value unless the team structure as well as the report structure is aligned to support
group framing. This study also shows that people have difficulties appreciating the value
of accounting and team structures as a means of affecting performance. Finally, the study
provides additional insight into the role that accounting can play in promoting informal
control within a severe control environment that can undermine standard formal control
mechanisms. This function is consistent with Chua and Degeling's (1993, 292) notion of a
"moral mediator" in which accounting operates by changing "the ways in which people
see, think, talk and relate to others."
This study also demonstrates that accounting can play an important role in governing
nonroutine process improvement tasks, which are becoming increasingly common in prac-
tice (restructuring, reengineering. and rapid, as opposed to incremental, adoption of process-
level technologies such as just-in-time operations). In particular, this study shows that ac-
counting structure is a powerful tool for shaping the informal "rules of the game" that
emerge within cross-functional teams that perform nonrouline tasks. In contrast, the man-
agement accounting literature has focused almost exclusively on routine or gradually
changing task settings (Bimberg 1998. 30-31).
A practical implication of this study is that when cross-functional teams are used to
plan process improvements, it is important to use "process-levei" accounting report struc-
ture together with "face-to-face" team structure. The study also finds support for the notion
that people need to be better educated about both the severity of the free-rider problem and
about the potential benefits of using accounting structure to resolve this problem.
Several interesting questions await future research. This study focused on the task of
planning process improvement initiatives within cross-functional teams. Follow-on studies

The Accounting Review, October 2004

The Effect of Accounting Report Structure and Team Structure on Performance 1177

could examine the role of accounting during subsequent phases of cross-functional tasks,
which may include participative negotiations and implementation.

APPENDIX
POST-EXPERIMENTAL INSTRUMENTS
Except where noted, each of the following items were measured on a nine-point scale
anchored by Strongly Disagree versus Strongly Agree.''

Trust
1. How many units do you predict that other team members will invest (range 0-6
units)? (Note that a $3 bonus was offered for a correct prediction).
2. I consider the members of my team to be trustworthy.
3. I am confident that the members of my team would not choose actions that would
harm me.
4. I expect that other team members considered my interests when they made their
investment choice.

Group Identification
1. I am glad to belong to this team.
2. I think this team will work well together.
3. I identify with this team.
4. I do not fit in well with the other members of this team (reverse scored).
5. I do not consider my team to be important (reverse scored).
6. I feel strong ties to this team.
7. I like my team members.

Collectivism
1. I think the right thing to do was to invest two units in my team's project.
2. I think it would be unfair if other managers choose to invest 0 units on the team
project.
3. I felt that I had a duty to invest my units of resources in the team project.
4. How much would you hypothetically pay to sanction other team members if they
choose to invest zero (range $0 to $3)?

REFERENCES
Abdel-khalik, A. R,. and E. J. Lusk. 1974. Transfer pricing—A synthesis. The Accounting; Review:
8-23.
Anderson. S. W., J. W. Hesford, and S. M. Young. 2002. Factors influencing the performance of
activity based costing teams: A field study of ABC model development time in the automobile
industry. Accounting, Organizations ami Society 37: 195-211.
Antle. R.. and G. Eppen. 1985. Capital rationing and organizational slack in capital budgeting. Man-
agement Science'i: 163-174.

Consistent with the experimental design, the words "team" or "team project" were only used in the face-io-
I'ace team condition and [his word was replaced by the term "joint project group" in the distributed leam
condition.

The Accounting Review, October 2004

1178 Rowe

Ashfoith, B. E., G. E. Kreiner, and M. Fugate. 2000. Alt in a day's work: Boundaries and micro role
transitions. Academy of Managemeni Review 25: 472-491.
Atkinson. A. A., R. Balakrishnan, P. Booth, J. M. Cote, T. Groot, T. Malmi, H. E. Roberts, and A.
Wu. 1997. New directions in management accounting research. Journal of Management Ac-
counting Research 9: 79-108.
Bimberg, J. G., and C. Snodgrass. 1988. Culture and control: A field study. Accounting, Organizations
and Society 13: 447^64.
. 1998. Some reflections on the evolution of organizational control. Behavioral Research in
Accounting 10: 27-46.
Blount S.. and R. P. Larrick. 2000. Framing the game: Examining frame choice in bargaining. Or-
ganizational Behavior and Human Decision Processes 81: 43-71.
Boylan, S. J., and G. B. Sprinkle. 2001. Experimental evidence on the relation between tax rates and
compliance: The effect of earned versus endowed income. The Journal of the American Taxation
Association 23: 75-91.
Briers. M.. and W. F. Chua. 2001. The role of actor-networks and boundary objects in management
accounting change: A field study of an implementation of activity-based costing. Accouniing.
Organizations and Society 26: 237-269.
Buckless, F. A., and S. P. Ravenscroft. 1990. Contrast coding: A refinement of ANOVA in behavioral
analysis. The Accouniing Review 65: 933-945.
Bushman R.. R. Indjejikian. and A. Smith. 1995. Aggregate performance measures in business unit
manager compensation: The role of intrafirm interdependencies. Journal of Accounting Research
33 (Supplement): 101-128.
Chua, W.. and P. Degeting. 1993. Interrogating and accounting-based intervention on three axes:
Instrumental, moral and aesthetic. Accounting, Organizations and Society 18: 291-318.
Cookson. R. 2000. Framing effects in public goods experiments. Experimental Economics 3: 55-79.
Dawes, R. 1980. Social dilemmas. Annual Review of Psychology 31: 169-193.
Denison, D.. S. Hart, and J. Kahn. 1996. From chimneys to cross-functional teams: Developing and
validating a diagnostic model. Academy of Management Journal 39: 1005-1023.
Donaldson, L. 2001. The Contingency Theory of Organizations. Thousand Oaks, CA: Sage.
Drake, A. R., S. F. Haka, and S. P. Ravenscroft. 1999. Cost system and incentive structure effects on
innovation, efficiency, and profitability in teams. The Accounting Review 74: 323-346.
Elster, J. 1989. The Cement of Society, New York, NY: Cambridge University Press.
Evans. J. H., R. L. Hannan, R. Krishnan. and D. V. Moser. 2001. Honesty in managerial reporting.
The Accounting Review 76: 537-559.
, Y. Hwang, and H. J. Nagarajan. 2001. Management control and hospital cost reduction:
Additional evidence. Journal of Accounting and Public Policy 20: 73-88.
Eiske. S. T. and S. E. Taylor. 1991. Social Cognition. 2nd edition. New York, NY: McGraw-Hill.
Gaertner, L., C. Sedikides, J. L. Vevea. and J. Iuzzini. 2002. The "I," the "we" and the "when": A
meta-analysis of motivational primacy in self-definition. Journal of Personality and Social Psy-
chology S3: 574-591.
Hays, W. L. 1994. Statistics. 5th edition. Fort Worth, TX: Harcourt Brace.
Hinkle, S., L. A. Taylor, D. L. Fox-Cardamone, and K. F. Crook. 1989. Intragroup identification and
intergroup differentiation: A multicomponent approach. British Journal of Social Psychology
28: 305-317.
Holmbeck, G. N. 1997. Toward terminological, conceptual, and statistical clarity in the study of
mediators and moderators. Journal of Consulting and Clinical Psychology 65: 599-610.
Holmstrom. B. 1982. Moral hazard in teams. Bell Journal of Economics 13: 324-340.
Ittner, C. D., and D. F. Larcker. 1995. Total quality management and the choice of information and
reward systems. Journal of Accounting Research 33: 1-34.
Jensen, M. C . and W. H. Meckling. 1995. Specific and general knowledge, and organizational struc-
ture. Journal of Applied Corporate Finance 8: 4-18.

The Accounting Review. October 2004

The Effect of Accounting Report Structure and Team Structure on Performance 1179

Joyce, W. E, V. E. McGee, and J. W. Slocum, Jr. 1997. Designing lateral organizations: An analysis
of the benefits, costs, and enablers of nonhierarchical organizational forms. Decision Sciences
28: 1-25.
Kachelmeier, S. J., and M. Shehata. 1997. Internal auditing and voluntary cooperation in firms: A
cross-cultural experiment. The Accounting Review 72: 407-432.
Kaplan, R. S., and D. P Norton. 1996. The Balanced Scorecard: Translating Strategy into Action.
Boston, MA: Harvard Business School Press.
Kenny. D. A., and C. Judd. 1996. A general procedure for the estimation of interdependence. Psy-
chological Bulletin 119: 138-147.
Kiesler, S., and J. N. Cummings. 2002. What do we know about proximity and distance in work
groups? A legacy of research. In Distributed Work, edited by P. J. Hinds, and S. Kiesler. 5 7 -
80. Cambridge, MA: The MIT Press.
Kilmann. R. 1983. The cost of organization structure: Dispelling the myths of independent divisions
and organization-wide decision making. Accounting. Organizations and Society 8: 341-357.
Komorita, S.. and C. D. Parks. 1995. Interpersonal relations: Mixed motive interaction. Annual Review
of Psychology 46: 183-207.
Kramer, R. 1999. Trust and distrust in organizations: Emerging perspectives, enduring questions.
Annual Review of Psychology 50: 569-598.
Kwok. W. C. C . and D. J. Sharp. 1998. A review of construct measurement issues in behavioral
accounting research. Journal of Accounting Literature 17: 137-174.
Ledyard, J. 1995. Public goods: A survey of experimental research. In The Handbook of Experimental
Economics, edited by J. H. Kagel, and A. E. Roth. Princeton, NJ: Princeton University Press.
Lind, E.. and T Tyler. 1988. The Social Psychology of Procedural Justice. New York, NY: Plenum
Press.
Luft. J., and M. D. Shields. 2003. Mapping management accounting: Graphics and guidelines for
theory-consistent empirical research. Accounting. Organizations and Society 28: 169-249.
Luhmann, N. 1995. Social systems. Stanford, CA: Stanford University Press.
Messick, D. M. 1999. Alternative logics for decision making in social settings. Journal of Economic
Behavior and Organization 39: 11-28.
Meyerson, D., K. Weick, and R. Kramer. 1995. Swift trust in temporary groups. In Trust in Organi-
zations, edited by R. M. Kramer, and T. R. Tyler. Thousand Oaks, CA: Sage.
Milgrom, P., and J. Roberts. 1992. Economics Organization and Management. Englewood Cliffs, NJ:
Prentice Hall.
Minsky. M. 1975. A framework for representing knowledge. In The Psychology of Computer Vision.
edited by F H. Winston, 211-280. New York. NY: McGraw-Hill.
Pick, J. 1971. Is responsibility accounting irresponsible? The New York Certified Public Accountant
(July): 487-494.
Ross, E.. and R. E. Nisbett. 1991. The Person and the Situation: Perspectives of Social Psychology.
Philadelphia. PA: Temple University Press.
Scott. T. and P. Tiessen. 1999. Performance measurement and managerial teams. Accounting. Orga-
nizations and Society 24: 263-285.
Shields. M. D. 1997. Research in management accounting by North Americans in the 1990s. Journal
of Management Accounting Research 9: 3-61.
Siegel, G., and J. E. Sorensen. 1999. Counting More. Counting Less: Transformations in the Man-
agement Accounting Profession. Montvale. NJ: Institute of Management Accountants.
Simon, H. A., H. Guetzkow, G. Kozmetsky, and G. Tyndall. 1954. Centralization versus Decentral-
ization in Organizing the Controller's Department. New York, NY: Controllership Foundation.
Slovic, P. 1993. Perceived risk, trust, and democracy. Risk Analysis 13: 675-682.
Sunder, S. 2002. Management control, expectations, common knowledge, and culture. Journal of
Management Accounting Research 14: 173-187.
Teece, D. J. 1996. Firm organization, industrial structure, and technological innovation. Journal of
Economic Behavior & Organization 31: 193-225.

The Accounting Review. October 2004

1180 Rowe

Towry, K. L. 2003. Control in a teamwork environment—The impact of social ties on the effectiveness
of mutual monitoring contracts. The Accounting Review 78: 1069-1095.
Triandis. H. 1995. Individualism and Collectivism. Boulder. CO: Westview.
Waller, W. S., and R. A. Bishop. 1990. An experimental study of incentive pay schemes, communi-
cation, and intratirm resource allocation. The Accounting Review 65: 812-836.
Weber, R., C. Camerer, Y. Rottenstreich, and M. Knez. 2001. The illusion of leadership: Misattribution
of cause in coordination games. Organization Science 12: 582-598.
Williamson, O. E. 1975. Markets and Hierarchies: Analysis and Antitrust Implications. New York,
NY: Free Press.
Wruck, K. H.. and M. C. Jensen. 1994. Science, specific knowledge, and total quality management.
Journal of Accounting and Economics 18: 247-287.
Young. M.. J. Fisher, and T. Lindquist. 1993. The effects of intergroup competition and intragroup
cooperation on slack and output in a manufacturing setting. The Accounting Review 68: 4 6 6 -
481.
Zand, D. E. 1997. The Leadership Triad: Knowledge. Tru.st, and Power. New York, NY: Oxford
University Press.

The Accounting Review. October 2004